llvm-project/llvm/test/Transforms/InstCombine/xor.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1174 lines
32 KiB
LLVM
Raw Normal View History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -instcombine -S | FileCheck %s
@G1 = global i32 0
@G2 = global i32 0
declare void @use(i8)
define i1 @test0(i1 %A) {
; CHECK-LABEL: @test0(
; CHECK-NEXT: ret i1 [[A:%.*]]
;
%B = xor i1 %A, false
ret i1 %B
}
define i32 @test1(i32 %A) {
; CHECK-LABEL: @test1(
; CHECK-NEXT: ret i32 [[A:%.*]]
;
%B = xor i32 %A, 0
ret i32 %B
}
define i1 @test2(i1 %A) {
; CHECK-LABEL: @test2(
; CHECK-NEXT: ret i1 false
;
%B = xor i1 %A, %A
ret i1 %B
}
define i32 @test3(i32 %A) {
; CHECK-LABEL: @test3(
; CHECK-NEXT: ret i32 0
;
%B = xor i32 %A, %A
ret i32 %B
}
define i32 @test4(i32 %A) {
; CHECK-LABEL: @test4(
; CHECK-NEXT: ret i32 -1
;
%NotA = xor i32 -1, %A
%B = xor i32 %A, %NotA
ret i32 %B
}
define i32 @test5(i32 %A) {
; CHECK-LABEL: @test5(
; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[A:%.*]], -124
; CHECK-NEXT: ret i32 [[TMP1]]
;
%t1 = or i32 %A, 123
%r = xor i32 %t1, 123
ret i32 %r
}
define i8 @test6(i8 %A) {
; CHECK-LABEL: @test6(
; CHECK-NEXT: ret i8 [[A:%.*]]
;
%B = xor i8 %A, 17
%C = xor i8 %B, 17
ret i8 %C
}
define i32 @test7(i32 %A, i32 %B) {
; CHECK-LABEL: @test7(
; CHECK-NEXT: [[A1:%.*]] = and i32 [[A:%.*]], 7
; CHECK-NEXT: [[B1:%.*]] = and i32 [[B:%.*]], 128
; CHECK-NEXT: [[C11:%.*]] = or i32 [[A1]], [[B1]]
; CHECK-NEXT: ret i32 [[C11]]
;
%A1 = and i32 %A, 7
%B1 = and i32 %B, 128
%C1 = xor i32 %A1, %B1
ret i32 %C1
}
define i8 @test8(i1 %c) {
; CHECK-LABEL: @test8(
; CHECK-NEXT: br i1 [[C:%.*]], label [[FALSE:%.*]], label [[TRUE:%.*]]
; CHECK: True:
; CHECK-NEXT: ret i8 1
; CHECK: False:
; CHECK-NEXT: ret i8 3
;
%d = xor i1 %c, true
br i1 %d, label %True, label %False
True:
ret i8 1
False:
ret i8 3
}
define i1 @test9(i8 %A) {
; CHECK-LABEL: @test9(
; CHECK-NEXT: [[C:%.*]] = icmp eq i8 [[A:%.*]], 89
; CHECK-NEXT: ret i1 [[C]]
;
%B = xor i8 %A, 123
%C = icmp eq i8 %B, 34
ret i1 %C
}
define <2 x i1> @test9vec(<2 x i8> %a) {
; CHECK-LABEL: @test9vec(
; CHECK-NEXT: [[C:%.*]] = icmp eq <2 x i8> [[A:%.*]], <i8 89, i8 89>
; CHECK-NEXT: ret <2 x i1> [[C]]
;
%b = xor <2 x i8> %a, <i8 123, i8 123>
%c = icmp eq <2 x i8> %b, <i8 34, i8 34>
ret <2 x i1> %c
}
define i8 @test10(i8 %A) {
; CHECK-LABEL: @test10(
; CHECK-NEXT: [[B:%.*]] = and i8 [[A:%.*]], 3
; CHECK-NEXT: [[C1:%.*]] = or i8 [[B]], 4
; CHECK-NEXT: ret i8 [[C1]]
;
%B = and i8 %A, 3
%C = xor i8 %B, 4
ret i8 %C
}
define i8 @test11(i8 %A) {
; CHECK-LABEL: @test11(
; CHECK-NEXT: [[B:%.*]] = and i8 [[A:%.*]], -13
; CHECK-NEXT: [[TMP1:%.*]] = or i8 [[B]], 8
; CHECK-NEXT: ret i8 [[TMP1]]
;
%B = or i8 %A, 12
%C = xor i8 %B, 4
ret i8 %C
}
define i1 @test12(i8 %A) {
; CHECK-LABEL: @test12(
; CHECK-NEXT: [[C:%.*]] = icmp ne i8 [[A:%.*]], 4
; CHECK-NEXT: ret i1 [[C]]
;
%B = xor i8 %A, 4
%c = icmp ne i8 %B, 0
ret i1 %c
}
define <2 x i1> @test12vec(<2 x i8> %a) {
; CHECK-LABEL: @test12vec(
; CHECK-NEXT: [[C:%.*]] = icmp ne <2 x i8> [[A:%.*]], <i8 4, i8 4>
; CHECK-NEXT: ret <2 x i1> [[C]]
;
%b = xor <2 x i8> %a, <i8 4, i8 4>
%c = icmp ne <2 x i8> %b, zeroinitializer
ret <2 x i1> %c
}
define i32 @test18(i32 %A) {
; CHECK-LABEL: @test18(
; CHECK-NEXT: [[C:%.*]] = add i32 [[A:%.*]], 124
; CHECK-NEXT: ret i32 [[C]]
;
%B = xor i32 %A, -1
%C = sub i32 123, %B
ret i32 %C
}
define i32 @test19(i32 %A, i32 %B) {
; CHECK-LABEL: @test19(
; CHECK-NEXT: ret i32 [[B:%.*]]
;
%C = xor i32 %A, %B
%D = xor i32 %C, %A
ret i32 %D
}
define void @test20(i32 %A, i32 %B) {
; CHECK-LABEL: @test20(
; CHECK-NEXT: store i32 [[B:%.*]], i32* @G1, align 4
; CHECK-NEXT: store i32 [[A:%.*]], i32* @G2, align 4
; CHECK-NEXT: ret void
;
%t2 = xor i32 %B, %A
%t5 = xor i32 %t2, %B
%t8 = xor i32 %t5, %t2
store i32 %t8, i32* @G1
store i32 %t5, i32* @G2
ret void
}
define i32 @test22(i1 %X) {
; CHECK-LABEL: @test22(
; CHECK-NEXT: [[Z:%.*]] = zext i1 [[X:%.*]] to i32
; CHECK-NEXT: ret i32 [[Z]]
;
%Y = xor i1 %X, true
%Z = zext i1 %Y to i32
%Q = xor i32 %Z, 1
ret i32 %Q
}
; Look through a zext between xors.
define i32 @fold_zext_xor_sandwich(i1 %X) {
; CHECK-LABEL: @fold_zext_xor_sandwich(
; CHECK-NEXT: [[Z:%.*]] = zext i1 [[X:%.*]] to i32
; CHECK-NEXT: [[Q:%.*]] = xor i32 [[Z]], 3
; CHECK-NEXT: ret i32 [[Q]]
;
%Y = xor i1 %X, true
%Z = zext i1 %Y to i32
%Q = xor i32 %Z, 2
ret i32 %Q
}
define <2 x i32> @fold_zext_xor_sandwich_vec(<2 x i1> %X) {
; CHECK-LABEL: @fold_zext_xor_sandwich_vec(
; CHECK-NEXT: [[Z:%.*]] = zext <2 x i1> [[X:%.*]] to <2 x i32>
; CHECK-NEXT: [[Q:%.*]] = xor <2 x i32> [[Z]], <i32 3, i32 3>
; CHECK-NEXT: ret <2 x i32> [[Q]]
;
%Y = xor <2 x i1> %X, <i1 true, i1 true>
%Z = zext <2 x i1> %Y to <2 x i32>
%Q = xor <2 x i32> %Z, <i32 2, i32 2>
ret <2 x i32> %Q
}
define i1 @test23(i32 %a, i32 %b) {
; CHECK-LABEL: @test23(
; CHECK-NEXT: [[T4:%.*]] = icmp eq i32 [[B:%.*]], 0
; CHECK-NEXT: ret i1 [[T4]]
;
%t2 = xor i32 %b, %a
%t4 = icmp eq i32 %t2, %a
ret i1 %t4
}
define i1 @test24(i32 %c, i32 %d) {
; CHECK-LABEL: @test24(
; CHECK-NEXT: [[T4:%.*]] = icmp ne i32 [[D:%.*]], 0
; CHECK-NEXT: ret i1 [[T4]]
;
%t2 = xor i32 %d, %c
%t4 = icmp ne i32 %t2, %c
ret i1 %t4
}
define i32 @test25(i32 %g, i32 %h) {
; CHECK-LABEL: @test25(
; CHECK-NEXT: [[T4:%.*]] = and i32 [[H:%.*]], [[G:%.*]]
; CHECK-NEXT: ret i32 [[T4]]
;
%h2 = xor i32 %h, -1
%t2 = and i32 %h2, %g
%t4 = xor i32 %t2, %g
ret i32 %t4
}
define i32 @test27(i32 %b, i32 %c, i32 %d) {
; CHECK-LABEL: @test27(
; CHECK-NEXT: [[T6:%.*]] = icmp eq i32 [[B:%.*]], [[C:%.*]]
; CHECK-NEXT: [[T7:%.*]] = zext i1 [[T6]] to i32
; CHECK-NEXT: ret i32 [[T7]]
;
%t2 = xor i32 %d, %b
%t5 = xor i32 %d, %c
%t6 = icmp eq i32 %t2, %t5
%t7 = zext i1 %t6 to i32
ret i32 %t7
}
define i32 @test28(i32 %indvar) {
; CHECK-LABEL: @test28(
; CHECK-NEXT: [[T214:%.*]] = add i32 [[INDVAR:%.*]], 1
; CHECK-NEXT: ret i32 [[T214]]
;
%t7 = add i32 %indvar, -2147483647
%t214 = xor i32 %t7, -2147483648
ret i32 %t214
}
define <2 x i32> @test28vec(<2 x i32> %indvar) {
; CHECK-LABEL: @test28vec(
; CHECK-NEXT: [[T214:%.*]] = add <2 x i32> [[INDVAR:%.*]], <i32 1, i32 1>
; CHECK-NEXT: ret <2 x i32> [[T214]]
;
%t7 = add <2 x i32> %indvar, <i32 -2147483647, i32 -2147483647>
%t214 = xor <2 x i32> %t7, <i32 -2147483648, i32 -2147483648>
ret <2 x i32> %t214
}
define i32 @test28_sub(i32 %indvar) {
; CHECK-LABEL: @test28_sub(
; CHECK-NEXT: [[T214:%.*]] = sub i32 1, [[INDVAR:%.*]]
; CHECK-NEXT: ret i32 [[T214]]
;
%t7 = sub i32 -2147483647, %indvar
%t214 = xor i32 %t7, -2147483648
ret i32 %t214
}
define <2 x i32> @test28_subvec(<2 x i32> %indvar) {
; CHECK-LABEL: @test28_subvec(
; CHECK-NEXT: [[T214:%.*]] = sub <2 x i32> <i32 1, i32 1>, [[INDVAR:%.*]]
; CHECK-NEXT: ret <2 x i32> [[T214]]
;
%t7 = sub <2 x i32> <i32 -2147483647, i32 -2147483647>, %indvar
%t214 = xor <2 x i32> %t7, <i32 -2147483648, i32 -2147483648>
ret <2 x i32> %t214
}
define i32 @test29(i1 %C) {
; CHECK-LABEL: @test29(
; CHECK-NEXT: [[V:%.*]] = select i1 [[C:%.*]], i32 915, i32 113
; CHECK-NEXT: ret i32 [[V]]
;
%A = select i1 %C, i32 1000, i32 10
%V = xor i32 %A, 123
ret i32 %V
}
define <2 x i32> @test29vec(i1 %C) {
; CHECK-LABEL: @test29vec(
; CHECK-NEXT: [[V:%.*]] = select i1 [[C:%.*]], <2 x i32> <i32 915, i32 915>, <2 x i32> <i32 113, i32 113>
; CHECK-NEXT: ret <2 x i32> [[V]]
;
%A = select i1 %C, <2 x i32> <i32 1000, i32 1000>, <2 x i32> <i32 10, i32 10>
%V = xor <2 x i32> %A, <i32 123, i32 123>
ret <2 x i32> %V
}
define <2 x i32> @test29vec2(i1 %C) {
; CHECK-LABEL: @test29vec2(
; CHECK-NEXT: [[V:%.*]] = select i1 [[C:%.*]], <2 x i32> <i32 915, i32 2185>, <2 x i32> <i32 113, i32 339>
; CHECK-NEXT: ret <2 x i32> [[V]]
;
%A = select i1 %C, <2 x i32> <i32 1000, i32 2500>, <2 x i32> <i32 10, i32 30>
%V = xor <2 x i32> %A, <i32 123, i32 333>
ret <2 x i32> %V
}
define i32 @test30(i1 %which) {
; CHECK-LABEL: @test30(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 [[WHICH:%.*]], label [[FINAL:%.*]], label [[DELAY:%.*]]
; CHECK: delay:
; CHECK-NEXT: br label [[FINAL]]
; CHECK: final:
; CHECK-NEXT: [[A:%.*]] = phi i32 [ 915, [[ENTRY:%.*]] ], [ 113, [[DELAY]] ]
; CHECK-NEXT: ret i32 [[A]]
;
entry:
br i1 %which, label %final, label %delay
delay:
br label %final
final:
%A = phi i32 [ 1000, %entry ], [ 10, %delay ]
%value = xor i32 %A, 123
ret i32 %value
}
define <2 x i32> @test30vec(i1 %which) {
; CHECK-LABEL: @test30vec(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 [[WHICH:%.*]], label [[FINAL:%.*]], label [[DELAY:%.*]]
; CHECK: delay:
; CHECK-NEXT: br label [[FINAL]]
; CHECK: final:
; CHECK-NEXT: [[A:%.*]] = phi <2 x i32> [ <i32 915, i32 915>, [[ENTRY:%.*]] ], [ <i32 113, i32 113>, [[DELAY]] ]
; CHECK-NEXT: ret <2 x i32> [[A]]
;
entry:
br i1 %which, label %final, label %delay
delay:
br label %final
final:
%A = phi <2 x i32> [ <i32 1000, i32 1000>, %entry ], [ <i32 10, i32 10>, %delay ]
%value = xor <2 x i32> %A, <i32 123, i32 123>
ret <2 x i32> %value
}
define <2 x i32> @test30vec2(i1 %which) {
; CHECK-LABEL: @test30vec2(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 [[WHICH:%.*]], label [[FINAL:%.*]], label [[DELAY:%.*]]
; CHECK: delay:
; CHECK-NEXT: br label [[FINAL]]
; CHECK: final:
; CHECK-NEXT: [[A:%.*]] = phi <2 x i32> [ <i32 915, i32 2185>, [[ENTRY:%.*]] ], [ <i32 113, i32 339>, [[DELAY]] ]
; CHECK-NEXT: ret <2 x i32> [[A]]
;
entry:
br i1 %which, label %final, label %delay
delay:
br label %final
final:
%A = phi <2 x i32> [ <i32 1000, i32 2500>, %entry ], [ <i32 10, i32 30>, %delay ]
%value = xor <2 x i32> %A, <i32 123, i32 333>
ret <2 x i32> %value
}
; B ^ (B | A) --> A & ~B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @or_xor_commute1(i32 %p1, i32 %p2) {
; CHECK-LABEL: @or_xor_commute1(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[B]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[A]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = or i32 %b, %a
%r = xor i32 %b, %o
ret i32 %r
}
; B ^ (B | A) --> A & ~B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @or_xor_commute2(i32 %p1, i32 %p2) {
; CHECK-LABEL: @or_xor_commute2(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[B]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[A]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = or i32 %a, %b
%r = xor i32 %o, %b
ret i32 %r
}
; B ^ (B | A) --> A & ~B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @or_xor_commute3(i32 %p1, i32 %p2) {
; CHECK-LABEL: @or_xor_commute3(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[B]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[A]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = or i32 %b, %a
%r = xor i32 %o, %b
ret i32 %r
}
; B ^ (B | A) --> A & ~B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @or_xor_commute4(i32 %p1, i32 %p2) {
; CHECK-LABEL: @or_xor_commute4(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[B]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[A]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = or i32 %a, %b
%r = xor i32 %b, %o
ret i32 %r
}
define i32 @or_xor_extra_use(i32 %a, i32 %b, i32* %p) {
; CHECK-LABEL: @or_xor_extra_use(
; CHECK-NEXT: [[O:%.*]] = or i32 [[A:%.*]], [[B:%.*]]
; CHECK-NEXT: store i32 [[O]], i32* [[P:%.*]], align 4
; CHECK-NEXT: [[R:%.*]] = xor i32 [[O]], [[B]]
; CHECK-NEXT: ret i32 [[R]]
;
%o = or i32 %a, %b
store i32 %o, i32* %p
%r = xor i32 %b, %o
ret i32 %r
}
; B ^ (B & A) --> ~A & B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @and_xor_commute1(i32 %p1, i32 %p2) {
; CHECK-LABEL: @and_xor_commute1(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[A]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[B]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = and i32 %b, %a
%r = xor i32 %b, %o
ret i32 %r
}
; B ^ (B & A) --> ~A & B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @and_xor_commute2(i32 %p1, i32 %p2) {
; CHECK-LABEL: @and_xor_commute2(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[A]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[B]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = and i32 %a, %b
%r = xor i32 %o, %b
ret i32 %r
}
; B ^ (B & A) --> ~A & B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @and_xor_commute3(i32 %p1, i32 %p2) {
; CHECK-LABEL: @and_xor_commute3(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[A]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[B]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = and i32 %b, %a
%r = xor i32 %o, %b
ret i32 %r
}
; B ^ (B & A) --> ~A & B
; The division ops are here to thwart complexity-based canonicalization: all ops are binops.
define i32 @and_xor_commute4(i32 %p1, i32 %p2) {
; CHECK-LABEL: @and_xor_commute4(
; CHECK-NEXT: [[A:%.*]] = udiv i32 42, [[P1:%.*]]
; CHECK-NEXT: [[B:%.*]] = udiv i32 42, [[P2:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[A]], -1
; CHECK-NEXT: [[R:%.*]] = and i32 [[B]], [[TMP1]]
; CHECK-NEXT: ret i32 [[R]]
;
%a = udiv i32 42, %p1
%b = udiv i32 42, %p2
%o = and i32 %a, %b
%r = xor i32 %b, %o
ret i32 %r
}
define i32 @and_xor_extra_use(i32 %a, i32 %b, i32* %p) {
; CHECK-LABEL: @and_xor_extra_use(
; CHECK-NEXT: [[O:%.*]] = and i32 [[A:%.*]], [[B:%.*]]
; CHECK-NEXT: store i32 [[O]], i32* [[P:%.*]], align 4
; CHECK-NEXT: [[R:%.*]] = xor i32 [[O]], [[B]]
; CHECK-NEXT: ret i32 [[R]]
;
%o = and i32 %a, %b
store i32 %o, i32* %p
%r = xor i32 %b, %o
ret i32 %r
}
; (~X | C2) ^ C1 --> ((X & ~C2) ^ -1) ^ C1 --> (X & ~C2) ^ ~C1
; The extra use (store) is here because the simpler case
; may be transformed using demanded bits.
define i8 @xor_or_not(i8 %x, i8* %p) {
; CHECK-LABEL: @xor_or_not(
; CHECK-NEXT: [[NX:%.*]] = xor i8 [[X:%.*]], -1
; CHECK-NEXT: store i8 [[NX]], i8* [[P:%.*]], align 1
; CHECK-NEXT: [[TMP1:%.*]] = and i8 [[X]], -8
; CHECK-NEXT: [[R:%.*]] = xor i8 [[TMP1]], -13
; CHECK-NEXT: ret i8 [[R]]
;
%nx = xor i8 %x, -1
store i8 %nx, i8* %p
%or = or i8 %nx, 7
%r = xor i8 %or, 12
ret i8 %r
}
; Don't do this if the 'or' has extra uses.
define i8 @xor_or_not_uses(i8 %x, i8* %p) {
; CHECK-LABEL: @xor_or_not_uses(
; CHECK-NEXT: [[NX:%.*]] = xor i8 [[X:%.*]], -1
; CHECK-NEXT: [[OR:%.*]] = or i8 [[NX]], 7
; CHECK-NEXT: store i8 [[OR]], i8* [[P:%.*]], align 1
; CHECK-NEXT: [[R:%.*]] = xor i8 [[OR]], 12
; CHECK-NEXT: ret i8 [[R]]
;
%nx = xor i8 %x, -1
%or = or i8 %nx, 7
store i8 %or, i8* %p
%r = xor i8 %or, 12
ret i8 %r
}
; (~X & C2) ^ C1 --> ((X | ~C2) ^ -1) ^ C1 --> (X | ~C2) ^ ~C1
; The extra use (store) is here because the simpler case
; may be transformed using demanded bits.
define i8 @xor_and_not(i8 %x, i8* %p) {
; CHECK-LABEL: @xor_and_not(
; CHECK-NEXT: [[NX:%.*]] = xor i8 [[X:%.*]], -1
; CHECK-NEXT: store i8 [[NX]], i8* [[P:%.*]], align 1
; CHECK-NEXT: [[TMP1:%.*]] = or i8 [[X]], -43
; CHECK-NEXT: [[R:%.*]] = xor i8 [[TMP1]], -32
; CHECK-NEXT: ret i8 [[R]]
;
%nx = xor i8 %x, -1
store i8 %nx, i8* %p
%and = and i8 %nx, 42
%r = xor i8 %and, 31
ret i8 %r
}
; Don't do this if the 'and' has extra uses.
define i8 @xor_and_not_uses(i8 %x, i8* %p) {
; CHECK-LABEL: @xor_and_not_uses(
; CHECK-NEXT: [[NX:%.*]] = and i8 [[X:%.*]], 42
; CHECK-NEXT: [[AND:%.*]] = xor i8 [[NX]], 42
; CHECK-NEXT: store i8 [[AND]], i8* [[P:%.*]], align 1
; CHECK-NEXT: [[R:%.*]] = xor i8 [[NX]], 53
; CHECK-NEXT: ret i8 [[R]]
;
%nx = xor i8 %x, -1
%and = and i8 %nx, 42
store i8 %and, i8* %p
%r = xor i8 %and, 31
ret i8 %r
}
; The tests 39-47 are related to the canonicalization:
; %notx = xor i32 %x, -1
; %cmp = icmp sgt i32 %notx, %y
; %smax = select i1 %cmp, i32 %notx, i32 %y
; %res = xor i32 %smax, -1
; =>
; %noty = xor i32 %y, -1
; %cmp2 = icmp slt %x, %noty
; %res = select i1 %cmp2, i32 %x, i32 %noty
;
; Same transformations is valid for smin/umax/umin.
define i32 @test39(i32 %x) {
; CHECK-LABEL: @test39(
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[X:%.*]], 255
[InstCombine] Process newly inserted instructions in the correct order InstCombine operates on the basic premise that the operands of the currently processed instruction have already been simplified. It achieves this by pushing instructions to the worklist in reverse program order, so that instructions are popped off in program order. The worklist management in the main combining loop also makes sure to uphold this invariant. However, the same is not true for all the code that is performing manual worklist management. The largest problem (addressed in this patch) are instructions inserted by InstCombine's IRBuilder. These will be pushed onto the worklist in order of insertion (generally matching program order), which means that a) the users of the original instruction will be visited first, as they are pushed later in the main loop and b) the newly inserted instructions will be visited in reverse program order. This causes a number of problems: First, folds operate on instructions that have not had their operands simplified, which may result in optimizations being missed (ran into this in https://reviews.llvm.org/D72048#1800424, which was the original motivation for this patch). Additionally, this increases the amount of folds InstCombine has to perform, both within one iteration, and by increasing the number of total iterations. This patch addresses the issue by adding a Worklist.AddDeferred() method, which is used for instructions inserted by IRBuilder. These will only be added to the real worklist after the combine finished, and in reverse order, so they will end up processed in program order. I should note that the same should also be done to nearly all other uses of Worklist.Add(), but I'm starting with just this occurrence, which has by far the largest test fallout. Most of the test changes are due to https://bugs.llvm.org/show_bug.cgi?id=44521 or other cases where we don't canonicalize something. These are neutral. One regression has been addressed in D73575 and D73647. The remaining regression in an shl+sdiv fold can't really be fixed without dropping another transform, but does not seem particularly problematic in the first place. Differential Revision: https://reviews.llvm.org/D73411
2020-01-02 00:56:37 +08:00
; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[X]], i32 255
; CHECK-NEXT: ret i32 [[TMP2]]
;
%1 = xor i32 %x, -1
%2 = icmp sgt i32 %1, -256
%3 = select i1 %2, i32 %1, i32 -256
%res = xor i32 %3, -1
ret i32 %res
}
define i32 @test40(i32 %x, i32 %y) {
; CHECK-LABEL: @test40(
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[Y:%.*]], -1
; CHECK-NEXT: [[TMP2:%.*]] = icmp sgt i32 [[TMP1]], [[X:%.*]]
; CHECK-NEXT: [[RES:%.*]] = select i1 [[TMP2]], i32 [[X]], i32 [[TMP1]]
; CHECK-NEXT: ret i32 [[RES]]
;
%notx = xor i32 %x, -1
%cmp1 = icmp sgt i32 %notx, %y
%smax = select i1 %cmp1, i32 %notx, i32 %y
%res = xor i32 %smax, -1
ret i32 %res
}
define i32 @test41(i32 %x, i32 %y) {
; CHECK-LABEL: @test41(
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[Y:%.*]], -1
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[TMP1]], [[X:%.*]]
; CHECK-NEXT: [[RES:%.*]] = select i1 [[TMP2]], i32 [[X]], i32 [[TMP1]]
; CHECK-NEXT: ret i32 [[RES]]
;
%notx = xor i32 %x, -1
%cmp1 = icmp slt i32 %notx, %y
%smin = select i1 %cmp1, i32 %notx, i32 %y
%res = xor i32 %smin, -1
ret i32 %res
}
define i32 @test42(i32 %x, i32 %y) {
; CHECK-LABEL: @test42(
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[Y:%.*]], -1
; CHECK-NEXT: [[TMP2:%.*]] = icmp ugt i32 [[TMP1]], [[X:%.*]]
; CHECK-NEXT: [[RES:%.*]] = select i1 [[TMP2]], i32 [[X]], i32 [[TMP1]]
; CHECK-NEXT: ret i32 [[RES]]
;
%notx = xor i32 %x, -1
%cmp1 = icmp ugt i32 %notx, %y
%umax = select i1 %cmp1, i32 %notx, i32 %y
%res = xor i32 %umax, -1
ret i32 %res
}
define i32 @test43(i32 %x, i32 %y) {
; CHECK-LABEL: @test43(
; CHECK-NEXT: [[TMP1:%.*]] = xor i32 [[Y:%.*]], -1
; CHECK-NEXT: [[TMP2:%.*]] = icmp ult i32 [[TMP1]], [[X:%.*]]
; CHECK-NEXT: [[RES:%.*]] = select i1 [[TMP2]], i32 [[X]], i32 [[TMP1]]
; CHECK-NEXT: ret i32 [[RES]]
;
%notx = xor i32 %x, -1
%cmp1 = icmp ult i32 %notx, %y
%umin = select i1 %cmp1, i32 %notx, i32 %y
%res = xor i32 %umin, -1
ret i32 %res
}
define i32 @test44(i32 %x, i32 %y) {
; CHECK-LABEL: @test44(
; CHECK-NEXT: [[TMP1:%.*]] = sub i32 -4, [[Y:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = icmp ugt i32 [[TMP1]], [[X:%.*]]
; CHECK-NEXT: [[RES:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 [[X]]
; CHECK-NEXT: ret i32 [[RES]]
;
%z = add i32 %y, 3 ; thwart complexity-based canonicalization
%notx = xor i32 %x, -1
%cmp1 = icmp ult i32 %z, %notx
%umin = select i1 %cmp1, i32 %z, i32 %notx
%res = xor i32 %umin, -1
ret i32 %res
}
define i32 @test45(i32 %x, i32 %y) {
; CHECK-LABEL: @test45(
[InstCombine] Process newly inserted instructions in the correct order InstCombine operates on the basic premise that the operands of the currently processed instruction have already been simplified. It achieves this by pushing instructions to the worklist in reverse program order, so that instructions are popped off in program order. The worklist management in the main combining loop also makes sure to uphold this invariant. However, the same is not true for all the code that is performing manual worklist management. The largest problem (addressed in this patch) are instructions inserted by InstCombine's IRBuilder. These will be pushed onto the worklist in order of insertion (generally matching program order), which means that a) the users of the original instruction will be visited first, as they are pushed later in the main loop and b) the newly inserted instructions will be visited in reverse program order. This causes a number of problems: First, folds operate on instructions that have not had their operands simplified, which may result in optimizations being missed (ran into this in https://reviews.llvm.org/D72048#1800424, which was the original motivation for this patch). Additionally, this increases the amount of folds InstCombine has to perform, both within one iteration, and by increasing the number of total iterations. This patch addresses the issue by adding a Worklist.AddDeferred() method, which is used for instructions inserted by IRBuilder. These will only be added to the real worklist after the combine finished, and in reverse order, so they will end up processed in program order. I should note that the same should also be done to nearly all other uses of Worklist.Add(), but I'm starting with just this occurrence, which has by far the largest test fallout. Most of the test changes are due to https://bugs.llvm.org/show_bug.cgi?id=44521 or other cases where we don't canonicalize something. These are neutral. One regression has been addressed in D73575 and D73647. The remaining regression in an shl+sdiv fold can't really be fixed without dropping another transform, but does not seem particularly problematic in the first place. Differential Revision: https://reviews.llvm.org/D73411
2020-01-02 00:56:37 +08:00
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i32 [[Y:%.*]], [[X:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[Y]], i32 [[X]]
; CHECK-NEXT: ret i32 [[TMP2]]
;
%z = xor i32 %y, -1
%notx = xor i32 %x, -1
%cmp1 = icmp ult i32 %z, %notx
%umin = select i1 %cmp1, i32 %z, i32 %notx
%res = xor i32 %umin, -1
ret i32 %res
}
; Check that we work with splat vectors also.
define <4 x i32> @test46(<4 x i32> %x) {
; CHECK-LABEL: @test46(
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt <4 x i32> [[X:%.*]], <i32 255, i32 255, i32 255, i32 255>
; CHECK-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[X]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>
; CHECK-NEXT: ret <4 x i32> [[TMP2]]
;
%1 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 -1, i32 -1>
%2 = icmp sgt <4 x i32> %1, <i32 -256, i32 -256, i32 -256, i32 -256>
%3 = select <4 x i1> %2, <4 x i32> %1, <4 x i32> <i32 -256, i32 -256, i32 -256, i32 -256>
%4 = xor <4 x i32> %3, <i32 -1, i32 -1, i32 -1, i32 -1>
ret <4 x i32> %4
}
; Test case when select pattern has more than one use.
define i32 @test47(i32 %x, i32 %y, i32 %z) {
; CHECK-LABEL: @test47(
; CHECK-NEXT: [[NOTX:%.*]] = xor i32 [[X:%.*]], -1
; CHECK-NEXT: [[CMP1:%.*]] = icmp ugt i32 [[NOTX]], [[Y:%.*]]
; CHECK-NEXT: [[UMAX:%.*]] = select i1 [[CMP1]], i32 [[NOTX]], i32 [[Y]]
; CHECK-NEXT: [[UMIN:%.*]] = xor i32 [[UMAX]], -1
; CHECK-NEXT: [[ADD:%.*]] = add i32 [[UMAX]], [[Z:%.*]]
; CHECK-NEXT: [[RES:%.*]] = mul i32 [[ADD]], [[UMIN]]
; CHECK-NEXT: ret i32 [[RES]]
;
%notx = xor i32 %x, -1
%cmp1 = icmp ugt i32 %notx, %y
%umax = select i1 %cmp1, i32 %notx, i32 %y
%umin = xor i32 %umax, -1
%add = add i32 %umax, %z
%res = mul i32 %umin, %add
ret i32 %res
}
define i32 @test48(i32 %x) {
; CHECK-LABEL: @test48(
; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[X:%.*]], 1
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[TMP1]], -1
; CHECK-NEXT: [[D:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 -1
; CHECK-NEXT: ret i32 [[D]]
;
%a = sub i32 -2, %x
%b = icmp sgt i32 %a, 0
%c = select i1 %b, i32 %a, i32 0
%d = xor i32 %c, -1
ret i32 %d
}
define <2 x i32> @test48vec(<2 x i32> %x) {
; CHECK-LABEL: @test48vec(
; CHECK-NEXT: [[TMP1:%.*]] = add <2 x i32> [[X:%.*]], <i32 1, i32 1>
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt <2 x i32> [[TMP1]], <i32 -1, i32 -1>
; CHECK-NEXT: [[D:%.*]] = select <2 x i1> [[TMP2]], <2 x i32> [[TMP1]], <2 x i32> <i32 -1, i32 -1>
; CHECK-NEXT: ret <2 x i32> [[D]]
;
%a = sub <2 x i32> <i32 -2, i32 -2>, %x
%b = icmp sgt <2 x i32> %a, zeroinitializer
%c = select <2 x i1> %b, <2 x i32> %a, <2 x i32> zeroinitializer
%d = xor <2 x i32> %c, <i32 -1, i32 -1>
ret <2 x i32> %d
}
define i32 @test49(i32 %x) {
; CHECK-LABEL: @test49(
; CHECK-NEXT: [[TMP1:%.*]] = sub i32 1, [[X:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = icmp sgt i32 [[TMP1]], 0
; CHECK-NEXT: [[D:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
; CHECK-NEXT: ret i32 [[D]]
;
%a = add i32 %x, -2
%b = icmp slt i32 %a, -1
%c = select i1 %b, i32 %a, i32 -1
%d = xor i32 %c, -1
ret i32 %d
}
define <2 x i32> @test49vec(<2 x i32> %x) {
; CHECK-LABEL: @test49vec(
; CHECK-NEXT: [[TMP1:%.*]] = sub <2 x i32> <i32 1, i32 1>, [[X:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = icmp sgt <2 x i32> [[TMP1]], zeroinitializer
; CHECK-NEXT: [[D:%.*]] = select <2 x i1> [[TMP2]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
; CHECK-NEXT: ret <2 x i32> [[D]]
;
%a = add <2 x i32> %x, <i32 -2, i32 -2>
%b = icmp slt <2 x i32> %a, <i32 -1, i32 -1>
%c = select <2 x i1> %b, <2 x i32> %a, <2 x i32> <i32 -1, i32 -1>
%d = xor <2 x i32> %c, <i32 -1, i32 -1>
ret <2 x i32> %d
}
define i32 @test50(i32 %x, i32 %y) {
; CHECK-LABEL: @test50(
; CHECK-NEXT: [[TMP1:%.*]] = sub i32 1, [[X:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = add i32 [[Y:%.*]], 1
[InstCombine] Process newly inserted instructions in the correct order InstCombine operates on the basic premise that the operands of the currently processed instruction have already been simplified. It achieves this by pushing instructions to the worklist in reverse program order, so that instructions are popped off in program order. The worklist management in the main combining loop also makes sure to uphold this invariant. However, the same is not true for all the code that is performing manual worklist management. The largest problem (addressed in this patch) are instructions inserted by InstCombine's IRBuilder. These will be pushed onto the worklist in order of insertion (generally matching program order), which means that a) the users of the original instruction will be visited first, as they are pushed later in the main loop and b) the newly inserted instructions will be visited in reverse program order. This causes a number of problems: First, folds operate on instructions that have not had their operands simplified, which may result in optimizations being missed (ran into this in https://reviews.llvm.org/D72048#1800424, which was the original motivation for this patch). Additionally, this increases the amount of folds InstCombine has to perform, both within one iteration, and by increasing the number of total iterations. This patch addresses the issue by adding a Worklist.AddDeferred() method, which is used for instructions inserted by IRBuilder. These will only be added to the real worklist after the combine finished, and in reverse order, so they will end up processed in program order. I should note that the same should also be done to nearly all other uses of Worklist.Add(), but I'm starting with just this occurrence, which has by far the largest test fallout. Most of the test changes are due to https://bugs.llvm.org/show_bug.cgi?id=44521 or other cases where we don't canonicalize something. These are neutral. One regression has been addressed in D73575 and D73647. The remaining regression in an shl+sdiv fold can't really be fixed without dropping another transform, but does not seem particularly problematic in the first place. Differential Revision: https://reviews.llvm.org/D73411
2020-01-02 00:56:37 +08:00
; CHECK-NEXT: [[TMP3:%.*]] = icmp sgt i32 [[TMP1]], [[TMP2]]
; CHECK-NEXT: [[E:%.*]] = select i1 [[TMP3]], i32 [[TMP1]], i32 [[TMP2]]
; CHECK-NEXT: ret i32 [[E]]
;
%a = add i32 %x, -2
%b = sub i32 -2, %y
%c = icmp slt i32 %a, %b
%d = select i1 %c, i32 %a, i32 %b
%e = xor i32 %d, -1
ret i32 %e
}
define <2 x i32> @test50vec(<2 x i32> %x, <2 x i32> %y) {
; CHECK-LABEL: @test50vec(
; CHECK-NEXT: [[TMP1:%.*]] = sub <2 x i32> <i32 1, i32 1>, [[X:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = add <2 x i32> [[Y:%.*]], <i32 1, i32 1>
[InstCombine] Process newly inserted instructions in the correct order InstCombine operates on the basic premise that the operands of the currently processed instruction have already been simplified. It achieves this by pushing instructions to the worklist in reverse program order, so that instructions are popped off in program order. The worklist management in the main combining loop also makes sure to uphold this invariant. However, the same is not true for all the code that is performing manual worklist management. The largest problem (addressed in this patch) are instructions inserted by InstCombine's IRBuilder. These will be pushed onto the worklist in order of insertion (generally matching program order), which means that a) the users of the original instruction will be visited first, as they are pushed later in the main loop and b) the newly inserted instructions will be visited in reverse program order. This causes a number of problems: First, folds operate on instructions that have not had their operands simplified, which may result in optimizations being missed (ran into this in https://reviews.llvm.org/D72048#1800424, which was the original motivation for this patch). Additionally, this increases the amount of folds InstCombine has to perform, both within one iteration, and by increasing the number of total iterations. This patch addresses the issue by adding a Worklist.AddDeferred() method, which is used for instructions inserted by IRBuilder. These will only be added to the real worklist after the combine finished, and in reverse order, so they will end up processed in program order. I should note that the same should also be done to nearly all other uses of Worklist.Add(), but I'm starting with just this occurrence, which has by far the largest test fallout. Most of the test changes are due to https://bugs.llvm.org/show_bug.cgi?id=44521 or other cases where we don't canonicalize something. These are neutral. One regression has been addressed in D73575 and D73647. The remaining regression in an shl+sdiv fold can't really be fixed without dropping another transform, but does not seem particularly problematic in the first place. Differential Revision: https://reviews.llvm.org/D73411
2020-01-02 00:56:37 +08:00
; CHECK-NEXT: [[TMP3:%.*]] = icmp sgt <2 x i32> [[TMP1]], [[TMP2]]
; CHECK-NEXT: [[E:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]]
; CHECK-NEXT: ret <2 x i32> [[E]]
;
%a = add <2 x i32> %x, <i32 -2, i32 -2>
%b = sub <2 x i32> <i32 -2, i32 -2>, %y
%c = icmp slt <2 x i32> %a, %b
%d = select <2 x i1> %c, <2 x i32> %a, <2 x i32> %b
%e = xor <2 x i32> %d, <i32 -1, i32 -1>
ret <2 x i32> %e
}
define i32 @test51(i32 %x, i32 %y) {
; CHECK-LABEL: @test51(
; CHECK-NEXT: [[TMP1:%.*]] = sub i32 -3, [[X:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = add i32 [[Y:%.*]], -3
[InstCombine] Process newly inserted instructions in the correct order InstCombine operates on the basic premise that the operands of the currently processed instruction have already been simplified. It achieves this by pushing instructions to the worklist in reverse program order, so that instructions are popped off in program order. The worklist management in the main combining loop also makes sure to uphold this invariant. However, the same is not true for all the code that is performing manual worklist management. The largest problem (addressed in this patch) are instructions inserted by InstCombine's IRBuilder. These will be pushed onto the worklist in order of insertion (generally matching program order), which means that a) the users of the original instruction will be visited first, as they are pushed later in the main loop and b) the newly inserted instructions will be visited in reverse program order. This causes a number of problems: First, folds operate on instructions that have not had their operands simplified, which may result in optimizations being missed (ran into this in https://reviews.llvm.org/D72048#1800424, which was the original motivation for this patch). Additionally, this increases the amount of folds InstCombine has to perform, both within one iteration, and by increasing the number of total iterations. This patch addresses the issue by adding a Worklist.AddDeferred() method, which is used for instructions inserted by IRBuilder. These will only be added to the real worklist after the combine finished, and in reverse order, so they will end up processed in program order. I should note that the same should also be done to nearly all other uses of Worklist.Add(), but I'm starting with just this occurrence, which has by far the largest test fallout. Most of the test changes are due to https://bugs.llvm.org/show_bug.cgi?id=44521 or other cases where we don't canonicalize something. These are neutral. One regression has been addressed in D73575 and D73647. The remaining regression in an shl+sdiv fold can't really be fixed without dropping another transform, but does not seem particularly problematic in the first place. Differential Revision: https://reviews.llvm.org/D73411
2020-01-02 00:56:37 +08:00
; CHECK-NEXT: [[TMP3:%.*]] = icmp slt i32 [[TMP1]], [[TMP2]]
; CHECK-NEXT: [[E:%.*]] = select i1 [[TMP3]], i32 [[TMP1]], i32 [[TMP2]]
; CHECK-NEXT: ret i32 [[E]]
;
%a = add i32 %x, 2
%b = sub i32 2, %y
%c = icmp sgt i32 %a, %b
%d = select i1 %c, i32 %a, i32 %b
%e = xor i32 %d, -1
ret i32 %e
}
define <2 x i32> @test51vec(<2 x i32> %x, <2 x i32> %y) {
; CHECK-LABEL: @test51vec(
; CHECK-NEXT: [[TMP1:%.*]] = sub <2 x i32> <i32 -3, i32 -3>, [[X:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = add <2 x i32> [[Y:%.*]], <i32 -3, i32 -3>
[InstCombine] Process newly inserted instructions in the correct order InstCombine operates on the basic premise that the operands of the currently processed instruction have already been simplified. It achieves this by pushing instructions to the worklist in reverse program order, so that instructions are popped off in program order. The worklist management in the main combining loop also makes sure to uphold this invariant. However, the same is not true for all the code that is performing manual worklist management. The largest problem (addressed in this patch) are instructions inserted by InstCombine's IRBuilder. These will be pushed onto the worklist in order of insertion (generally matching program order), which means that a) the users of the original instruction will be visited first, as they are pushed later in the main loop and b) the newly inserted instructions will be visited in reverse program order. This causes a number of problems: First, folds operate on instructions that have not had their operands simplified, which may result in optimizations being missed (ran into this in https://reviews.llvm.org/D72048#1800424, which was the original motivation for this patch). Additionally, this increases the amount of folds InstCombine has to perform, both within one iteration, and by increasing the number of total iterations. This patch addresses the issue by adding a Worklist.AddDeferred() method, which is used for instructions inserted by IRBuilder. These will only be added to the real worklist after the combine finished, and in reverse order, so they will end up processed in program order. I should note that the same should also be done to nearly all other uses of Worklist.Add(), but I'm starting with just this occurrence, which has by far the largest test fallout. Most of the test changes are due to https://bugs.llvm.org/show_bug.cgi?id=44521 or other cases where we don't canonicalize something. These are neutral. One regression has been addressed in D73575 and D73647. The remaining regression in an shl+sdiv fold can't really be fixed without dropping another transform, but does not seem particularly problematic in the first place. Differential Revision: https://reviews.llvm.org/D73411
2020-01-02 00:56:37 +08:00
; CHECK-NEXT: [[TMP3:%.*]] = icmp slt <2 x i32> [[TMP1]], [[TMP2]]
; CHECK-NEXT: [[E:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]]
; CHECK-NEXT: ret <2 x i32> [[E]]
;
%a = add <2 x i32> %x, <i32 2, i32 2>
%b = sub <2 x i32> <i32 2, i32 2>, %y
%c = icmp sgt <2 x i32> %a, %b
%d = select <2 x i1> %c, <2 x i32> %a, <2 x i32> %b
%e = xor <2 x i32> %d, <i32 -1, i32 -1>
ret <2 x i32> %e
}
define i4 @or_or_xor(i4 %x, i4 %y, i4 %z) {
; CHECK-LABEL: @or_or_xor(
; CHECK-NEXT: [[TMP1:%.*]] = xor i4 [[Z:%.*]], -1
; CHECK-NEXT: [[TMP2:%.*]] = xor i4 [[X:%.*]], [[Y:%.*]]
; CHECK-NEXT: [[R:%.*]] = and i4 [[TMP2]], [[TMP1]]
; CHECK-NEXT: ret i4 [[R]]
;
%o1 = or i4 %z, %x
%o2 = or i4 %z, %y
%r = xor i4 %o1, %o2
ret i4 %r
}
define i4 @or_or_xor_commute1(i4 %x, i4 %y, i4 %z) {
; CHECK-LABEL: @or_or_xor_commute1(
; CHECK-NEXT: [[TMP1:%.*]] = xor i4 [[Z:%.*]], -1
; CHECK-NEXT: [[TMP2:%.*]] = xor i4 [[X:%.*]], [[Y:%.*]]
; CHECK-NEXT: [[R:%.*]] = and i4 [[TMP2]], [[TMP1]]
; CHECK-NEXT: ret i4 [[R]]
;
%o1 = or i4 %x, %z
%o2 = or i4 %z, %y
%r = xor i4 %o1, %o2
ret i4 %r
}
define i4 @or_or_xor_commute2(i4 %x, i4 %y, i4 %z) {
; CHECK-LABEL: @or_or_xor_commute2(
; CHECK-NEXT: [[TMP1:%.*]] = xor i4 [[Z:%.*]], -1
; CHECK-NEXT: [[TMP2:%.*]] = xor i4 [[X:%.*]], [[Y:%.*]]
; CHECK-NEXT: [[R:%.*]] = and i4 [[TMP2]], [[TMP1]]
; CHECK-NEXT: ret i4 [[R]]
;
%o1 = or i4 %z, %x
%o2 = or i4 %y, %z
%r = xor i4 %o1, %o2
ret i4 %r
}
define <2 x i4> @or_or_xor_commute3(<2 x i4> %x, <2 x i4> %y, <2 x i4> %z) {
; CHECK-LABEL: @or_or_xor_commute3(
; CHECK-NEXT: [[TMP1:%.*]] = xor <2 x i4> [[Z:%.*]], <i4 -1, i4 -1>
; CHECK-NEXT: [[TMP2:%.*]] = xor <2 x i4> [[X:%.*]], [[Y:%.*]]
; CHECK-NEXT: [[R:%.*]] = and <2 x i4> [[TMP2]], [[TMP1]]
; CHECK-NEXT: ret <2 x i4> [[R]]
;
%o1 = or <2 x i4> %x, %z
%o2 = or <2 x i4> %y, %z
%r = xor <2 x i4> %o1, %o2
ret <2 x i4> %r
}
define i4 @or_or_xor_use1(i4 %x, i4 %y, i4 %z, i4* %p) {
; CHECK-LABEL: @or_or_xor_use1(
; CHECK-NEXT: [[O1:%.*]] = or i4 [[Z:%.*]], [[X:%.*]]
; CHECK-NEXT: store i4 [[O1]], i4* [[P:%.*]], align 1
; CHECK-NEXT: [[O2:%.*]] = or i4 [[Z]], [[Y:%.*]]
; CHECK-NEXT: [[R:%.*]] = xor i4 [[O1]], [[O2]]
; CHECK-NEXT: ret i4 [[R]]
;
%o1 = or i4 %z, %x
store i4 %o1, i4* %p
%o2 = or i4 %z, %y
%r = xor i4 %o1, %o2
ret i4 %r
}
define i4 @or_or_xor_use2(i4 %x, i4 %y, i4 %z, i4* %p) {
; CHECK-LABEL: @or_or_xor_use2(
; CHECK-NEXT: [[O1:%.*]] = or i4 [[Z:%.*]], [[X:%.*]]
; CHECK-NEXT: [[O2:%.*]] = or i4 [[Z]], [[Y:%.*]]
; CHECK-NEXT: store i4 [[O2]], i4* [[P:%.*]], align 1
; CHECK-NEXT: [[R:%.*]] = xor i4 [[O1]], [[O2]]
; CHECK-NEXT: ret i4 [[R]]
;
%o1 = or i4 %z, %x
%o2 = or i4 %z, %y
store i4 %o2, i4* %p
%r = xor i4 %o1, %o2
ret i4 %r
}
; PR32706 - https://bugs.llvm.org/show_bug.cgi?id=32706
; Pin an xor constant operand to -1 if possible because 'not' is better for SCEV and codegen.
define i32 @not_is_canonical(i32 %x, i32 %y) {
; CHECK-LABEL: @not_is_canonical(
; CHECK-NEXT: [[SUB:%.*]] = xor i32 [[X:%.*]], -1
; CHECK-NEXT: [[ADD:%.*]] = add i32 [[SUB]], [[Y:%.*]]
; CHECK-NEXT: [[MUL:%.*]] = shl i32 [[ADD]], 2
; CHECK-NEXT: ret i32 [[MUL]]
;
%sub = xor i32 %x, 1073741823
%add = add i32 %sub, %y
%mul = shl i32 %add, 2
ret i32 %mul
}
define i8 @not_shl(i8 %x) {
; CHECK-LABEL: @not_shl(
; CHECK-NEXT: [[TMP1:%.*]] = xor i8 [[X:%.*]], -1
; CHECK-NEXT: [[R:%.*]] = shl i8 [[TMP1]], 7
; CHECK-NEXT: ret i8 [[R]]
;
%a = shl i8 %x, 7
%r = xor i8 %a, 128
ret i8 %r
}
define <2 x i8> @not_shl_vec(<2 x i8> %x) {
; CHECK-LABEL: @not_shl_vec(
; CHECK-NEXT: [[TMP1:%.*]] = xor <2 x i8> [[X:%.*]], <i8 -1, i8 -1>
; CHECK-NEXT: [[R:%.*]] = shl <2 x i8> [[TMP1]], <i8 5, i8 5>
; CHECK-NEXT: ret <2 x i8> [[R]]
;
%a = shl <2 x i8> %x, <i8 5, i8 5>
%r = xor <2 x i8> %a, <i8 224, i8 224>
ret <2 x i8> %r
}
; negative test
define i8 @not_shl_extra_use(i8 %x) {
; CHECK-LABEL: @not_shl_extra_use(
; CHECK-NEXT: [[A:%.*]] = shl i8 [[X:%.*]], 7
; CHECK-NEXT: call void @use(i8 [[A]])
; CHECK-NEXT: [[R:%.*]] = xor i8 [[A]], -128
; CHECK-NEXT: ret i8 [[R]]
;
%a = shl i8 %x, 7
call void @use(i8 %a)
%r = xor i8 %a, 128
ret i8 %r
}
; negative test
define i8 @not_shl_wrong_const(i8 %x) {
; CHECK-LABEL: @not_shl_wrong_const(
; CHECK-NEXT: [[A:%.*]] = shl i8 [[X:%.*]], 6
; CHECK-NEXT: [[R:%.*]] = xor i8 [[A]], -128
; CHECK-NEXT: ret i8 [[R]]
;
%a = shl i8 %x, 6
%r = xor i8 %a, 128
ret i8 %r
}
define i8 @not_lshr(i8 %x) {
; CHECK-LABEL: @not_lshr(
; CHECK-NEXT: [[TMP1:%.*]] = xor i8 [[X:%.*]], -1
; CHECK-NEXT: [[R:%.*]] = lshr i8 [[TMP1]], 5
; CHECK-NEXT: ret i8 [[R]]
;
%a = lshr i8 %x, 5
%r = xor i8 %a, 7
ret i8 %r
}
define <2 x i8> @not_lshr_vec(<2 x i8> %x) {
; CHECK-LABEL: @not_lshr_vec(
; CHECK-NEXT: [[TMP1:%.*]] = xor <2 x i8> [[X:%.*]], <i8 -1, i8 -1>
; CHECK-NEXT: [[R:%.*]] = lshr <2 x i8> [[TMP1]], <i8 7, i8 7>
; CHECK-NEXT: ret <2 x i8> [[R]]
;
%a = lshr <2 x i8> %x, <i8 7, i8 7>
%r = xor <2 x i8> %a, <i8 1, i8 1>
ret <2 x i8> %r
}
; negative test
define i8 @not_lshr_extra_use(i8 %x) {
; CHECK-LABEL: @not_lshr_extra_use(
; CHECK-NEXT: [[A:%.*]] = lshr i8 [[X:%.*]], 5
; CHECK-NEXT: call void @use(i8 [[A]])
; CHECK-NEXT: [[R:%.*]] = xor i8 [[A]], 7
; CHECK-NEXT: ret i8 [[R]]
;
%a = lshr i8 %x, 5
call void @use(i8 %a)
%r = xor i8 %a, 7
ret i8 %r
}
; negative test
define i8 @not_lshr_wrong_const(i8 %x) {
; CHECK-LABEL: @not_lshr_wrong_const(
; CHECK-NEXT: [[A:%.*]] = lshr i8 [[X:%.*]], 5
; CHECK-NEXT: [[R:%.*]] = xor i8 [[A]], 3
; CHECK-NEXT: ret i8 [[R]]
;
%a = lshr i8 %x, 5
%r = xor i8 %a, 3
ret i8 %r
}
define i8 @ashr_not(i8 %x) {
; CHECK-LABEL: @ashr_not(
; CHECK-NEXT: [[N:%.*]] = ashr i8 [[X:%.*]], 5
; CHECK-NEXT: [[R:%.*]] = xor i8 [[N]], -1
; CHECK-NEXT: ret i8 [[R]]
;
%n = xor i8 %x, -1
%r = ashr i8 %n, 5
ret i8 %r
}
; Unlike the logicial shifts, 'not' is canonicalized after ashr.
define i8 @not_ashr(i8 %x) {
; CHECK-LABEL: @not_ashr(
; CHECK-NEXT: [[A:%.*]] = ashr i8 [[X:%.*]], 5
; CHECK-NEXT: [[R:%.*]] = xor i8 [[A]], -1
; CHECK-NEXT: ret i8 [[R]]
;
%a = ashr i8 %x, 5
%r = xor i8 %a, -1
ret i8 %r
}
define <2 x i8> @not_ashr_vec(<2 x i8> %x) {
; CHECK-LABEL: @not_ashr_vec(
; CHECK-NEXT: [[A:%.*]] = ashr <2 x i8> [[X:%.*]], <i8 7, i8 7>
; CHECK-NEXT: [[R:%.*]] = xor <2 x i8> [[A]], <i8 -1, i8 -1>
; CHECK-NEXT: ret <2 x i8> [[R]]
;
%a = ashr <2 x i8> %x, <i8 7, i8 7>
%r = xor <2 x i8> %a, <i8 -1, i8 -1>
ret <2 x i8> %r
}
define i8 @not_ashr_extra_use(i8 %x) {
; CHECK-LABEL: @not_ashr_extra_use(
; CHECK-NEXT: [[A:%.*]] = ashr i8 [[X:%.*]], 5
; CHECK-NEXT: call void @use(i8 [[A]])
; CHECK-NEXT: [[R:%.*]] = xor i8 [[A]], -1
; CHECK-NEXT: ret i8 [[R]]
;
%a = ashr i8 %x, 5
call void @use(i8 %a)
%r = xor i8 %a, -1
ret i8 %r
}
define i8 @not_ashr_wrong_const(i8 %x) {
; CHECK-LABEL: @not_ashr_wrong_const(
; CHECK-NEXT: [[A:%.*]] = ashr i8 [[X:%.*]], 5
; CHECK-NEXT: [[R:%.*]] = xor i8 [[A]], -2
; CHECK-NEXT: ret i8 [[R]]
;
%a = ashr i8 %x, 5
%r = xor i8 %a, -2
ret i8 %r
}